摘要
为了使用机器视觉实现对食用菌中发丝等杂质的自动检测,提出基于显著性特征的菌菇中杂质图像分割算法,该算法结合了Hessian灰度特征和Lab空间色彩特征,通过图像归一化、求Hessian矩阵、反向投影、取阈值分割出杂质图像。实验结果表明,该算法在光照不均匀条件下的识别率仍达到99.6%,可以用于工业化生产。
In order to achieve automatic recognition of hair impurities in edible mushrooms industry, a fingerprint image segmentation method based on impurity＇s distinguishing feature in mushrooms was proposed in this paper. This algorithm segments the impurity＇s image through the way of the image normalization, taking the Hessian matrix back projection, taking the threshold and combining the Hessian grayscale characteristics and Lab color space. These experimental results show that it also performs well and recongnition rate is up to 99. 6% in the case of nonuniform lighting conditions,which can be used in industrial production.
出处
《计算机科学》
CSCD
北大核心
2015年第B11期203-205,217,共4页
Computer Science
关键词
机器视觉
黑塞矩阵
反向投影
杂质检测
Machine vision, Hessian matrix, Back projection, Impurity recognition
